|
In order to realize the high-precision requirements of robotic multi-layer multi-pass welding and to improve the accuracy of weld bevel information recognition, a system based on laser vision for three-dimensional weld bevel recognition and reconstruction is established. Through the line structure optical sensor connected to the welding gun at the end of the welding robot, the welding seam is collected, and the noise generated by the reflections of the weldment and transmission interference is effectively reduced by threshold segmentation, adaptive selection of the region of interest, joint filtering processing, extraction of the center line and refinement of the collected data; Through the processed data still exists a small part of the existence of interference noise, affecting the subsequent recognition accuracy, the point-line projection method will be processed to obtain smooth image information; In its difference calculation, to obtain the feature point mutation information, to realize the accurate extraction of feature points; Through the transformation relationship between coordinate systems, the transformed data information is obtained, followed by computational solving to obtain the characteristic information of the 3D weld seam; The position calculation of the sensor's first frame of light bar information is carried out through the acquisition of the conversion relationship to scan at the optimal position, and the sensor and robot are controlled to acquire at the optimal parameters to obtain the highly reproducible 3D weld bevel's morphology. The experimental results show that the average error of bevel width and height after weld recognition is 0.1607mm and 0.1592mm, which meets the accuracy requirements of robot welding; Meanwhile, the reconstructed 3D weld bevel morphology has high reducibility, which provides a reference for realizing intelligent and high-precision autonomous welding.
|
